In [6]:
show_image(get_national_team())
import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
#Visiulazation
import matplotlib.pyplot as plt
#image processing
import cv2
#extracting zippped file
import tarfile
#systems
import os
print(os.listdir("../input/haarcascade-frontal-faces"))
print(os.listdir("../input/lfwpeople"))
['haarcascade_frontalface_default.xml', 'millitakim-isvecmaci.jpg'] ['pairsDevTest.txt', 'pairs.txt', 'pairsDevTrain.txt', 'lfwfunneled.tgz']
Go to Content Menu | Quick Links: 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8
In this study, face detection was performed from the images in the LFW data set.
Before face detection in LFW images, multiple face detection performed in single images.
Desired number of images are taken from the zip file in the LFW data set. Then the face detection was made from these images.
class FaceDetector():
def __init__(self,faceCascadePath):
self.faceCascade=cv2.CascadeClassifier(faceCascadePath)
def detect(self, image, scaleFactor=1.1,
minNeighbors=5,
minSize=(30,30)):
#function return rectangle coordinates of faces for given image
rects=self.faceCascade.detectMultiScale(image,
scaleFactor=scaleFactor,
minNeighbors=minNeighbors,
minSize=minSize)
return rects
#Frontal face of haar cascade loaded
frontal_cascade_path="../input/haarcascade-frontal-faces/haarcascade_frontalface_default.xml"
#Detector object created
fd=FaceDetector(frontal_cascade_path)
#An image contains faces, loaded
national_team_org=cv2.imread("../input/haarcascade-frontal-faces/millitakim-isvecmaci.jpg")
def get_national_team():
return np.copy(national_team_org)
def show_image(image):
plt.figure(figsize=(18,15))
#Before showing image, bgr color order transformed to rgb order
plt.imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
plt.xticks([])
plt.yticks([])
plt.show()
show_image(get_national_team())
def detect_face(image, scaleFactor, minNeighbors, minSize):
# face will detected in gray image
image_gray=cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
faces=fd.detect(image_gray,
scaleFactor=scaleFactor,
minNeighbors=minNeighbors,
minSize=minSize)
for x, y, w, h in faces:
#detected faces shown in color image
cv2.rectangle(image,(x,y),(x+w, y+h),(127, 255,0),3)
show_image(image)
national_team=get_national_team()
detect_face(image=national_team,
scaleFactor=1.9,
minNeighbors=3,
minSize=(30,30))
national_team=get_national_team()
#Let's play around function parameters
detect_face(image=national_team,
scaleFactor=1.3,
minNeighbors=3,
minSize=(30,30))
class FetchLFW:
def __init__(self, path):
self.path=path
def _initialize(self,dim):
self.dim_of_photo_gallery=dim
self.number_of_images=self.dim_of_photo_gallery*self.dim_of_photo_gallery
total_number_of_image=13233
self.random_face_indexes=np.arange(total_number_of_image)
np.random.shuffle(self.random_face_indexes)
self.n_random_face_indexes=self.random_face_indexes[:self.number_of_images]
def get_lfw_images(self,dim=5):
self._initialize(dim)
self.lfw_images=self._get_images()
return self.lfw_images
def _get_images(self):
image_list=[]
tar = tarfile.open(path, "r:gz")
counter=0
for tarinfo in tar:
tar.extract(tarinfo.name)
if tarinfo.name[-4:]==".jpg":
if counter in self.n_random_face_indexes:
image=cv2.imread(tarinfo.name, cv2.IMREAD_COLOR)
image=cv2.resize(image,None,fx=0.4, fy=0.4, interpolation=cv2.INTER_AREA)
image_list.append(np.array(image))
counter+=1
if tarinfo.isdir():
pass
else:
os.remove(tarinfo.name)
tar.close()
return np.array(image_list)
path="../input/lfwpeople/lfwfunneled.tgz"
fetchLFW=FetchLFW(path)
dimension=10
images=fetchLFW.get_lfw_images(dim=dimension)
print("LFW images have been fetched")
LFW images have been fetched
print(images.shape)
(100, 100, 100, 3)
def get_photo_gallery():
counter=0
himages=[]
vimages=[]
for i in range(dimension):
for j in range(dimension):
himages.append(images[counter])
counter+=1
himage=np.hstack((himages))
vimages.append(himage)
himages=[]
image_matrix=np.vstack((vimages))
return image_matrix
photo_gallery=get_photo_gallery()
print("photo_gallery:{}".format(photo_gallery.shape))
show_image(photo_gallery)
photo_gallery:(1000, 1000, 3)
frontal_cascade_path="../input/haarcascade-frontal-faces/haarcascade_frontalface_default.xml"
fd=FaceDetector(frontal_cascade_path)
face_counter=0
for image_org in images:
image_gray=cv2.cvtColor(image_org,cv2.COLOR_BGR2GRAY)
faceRect=fd.detect(image_gray,
scaleFactor=1.1,
minNeighbors=5,
minSize=(30,30))
#print("I found {} faces".format(len(faceRect)))
first_detection=False
for (x,y,w,h) in faceRect:
if first_detection==False:
face_counter+=1
cv2.rectangle(image_org,(x,y),(x+w,y+h),(127,255,0),2)
first_detection=True
else:
print("Second detection ignored in a image")
print("{} images have been scaned".format(dimension*dimension))
print("{} faces have been detected".format(face_counter))
100 images have been scaned 96 faces have been detected
photo_gallery=get_photo_gallery()
show_image(photo_gallery)
